Finally, I've got a couple of questions regarding the folks we're working with on doing
 the hand-count comparison with the scanner tally.

We've kicked around three general methods, which I'll describe. Can you get someone
to tell me how feasible each of the three are, how involved it would be to do each of
them, whether any would require additional work on your part, and how disruptive it
would be to the ballot counting process.

Easiest approach first: Hand count a precinct(s) or a race(s) in a precinct(s). It would
seem that once we got through counting a precinct - meaning the votes cast on election
day received from the precincts, and not an early vote or absentee vote, we could
compare the precincts tally from the statement of vote with the hand count whenever it
was completed. Minimal disruption.

Second approach: Taking random batches of 50 out of random precinct ballot boxes and
processing those in batches of 50. Is there a way of separating those batches of 50,
perhaps dropping them in a separate made up precinct, such as one called "hand count
comparison" precinct, and getting a tally for each batch to compare with a hand tally.
Batch one might represent the sample from precinct 1, batch 2 from precinct 13, etc.,
with the goal of seeing whether those precinct batches were tallied correctly. The two
questions I have are: "Can you develop a tally for a batch? Or could you at least get a
tally under the precinct report that says (say, in addition to Early Voting, Absentee
Voting and Election Day voting) "Sample Vote" that would be included in the precinct
count, yet discrete so that we could compare it against the hand count?

Third approach: A more scientifically rigorous approach that would involve some sort of
statistically valid random sample of the ballots being counted by a particular scanner.
The methodology would give some statistical certainty that a scanner was counting the
ballots correctly. Disregarding how the sample ballots were selected and pulled for
immediate purposes, this approach would seem to again rely on being able to tally the
batch or batches individually - as above - per machine. Again, I guess they could also be
labeled "sample vote" under the precinct reports, and assuming some record was kept of
which precincts were scanned by which machine, the total "sample vote" for each of
those precincts could be summed.

Let us know if these three approaches are feasible with the system, or if there is a better
way of going about this than I've suggested for a particular approach, or if there is a
method you'd recommend that's altogether different but accomplishes the same goal
while being easier/less complicated to do.

Tom